CN112270999B

CN112270999B - Epidemic early detection system and method based on big data and artificial intelligence

Info

Publication number: CN112270999B
Application number: CN202011066664.XA
Authority: CN
Inventors: 孙炜
Original assignee: Individual
Current assignee: Individual
Priority date: 2020-10-02
Filing date: 2020-10-02
Publication date: 2022-10-11
Anticipated expiration: 2040-10-02
Also published as: CN112270999A

Abstract

The invention provides a epidemic early detection system and a method based on big data and artificial intelligence, wherein the system comprises an intelligent data acquisition device and a data center; the intelligent data acquisition device acquires and transmits information data center; s1: collecting survey information, preliminarily analyzing and judging the tendency of the infectious disease, and forming a recommendation conclusion by the collected survey information and a judgment result and transmitting the recommendation conclusion to a data center; s2: the data center collects all the obtained suspected syndromes to form a global syndrome list, and the corresponding value of the suspected syndromes is a to form a list; s3: the data center judges whether the signs of the infectious diseases exist or not by using the syndrome data in the steps and constructs an infectious spatio-temporal network by using the spatio-temporal data; s4: judging the transmission path of the infectious disease through the infection space-time network obtained in the step S3 and carrying out infectious disease inspection and investigation; the beneficial effects of the invention are as follows: the method overcomes the defects of early epidemics investigation in the prior art and realizes objective and active epidemics investigation and judgment.

Description

Epidemic early-stage discovery system and method based on big data and artificial intelligence

Technical Field

The invention relates to the field of disease monitoring and judgment, in particular to an epidemic early-stage discovery system and method based on big data and artificial intelligence.

Background

The existing epidemic situation direct reporting system has some lags and defects in dealing with the emergent infectious diseases, and the key problem is that the system depends on manual recording and input and on subjective judgment and emotional factors, so that serious problems are caused to find the lagged property. Meanwhile, the system can only report to arouse attention of the supervision personnel, and then the epidemic disease investigation is carried out by the supervision personnel, so that the hysteresis of epidemic situation control is caused.

The artificial epidemic investigation has high cost, long period and high risk. Although some tracking control systems, such as health codes, are available, they rely entirely on human input and are too subjective. Individual systems, such as a healthy cloud, can make simple artificial intelligence inquiries. But infectivity was not judged.

Some tools and products for judging and analyzing the symptom indexes of a single person exist in the industry, but the disease is difficult to judge and the infectious disease is difficult to judge only through symptoms and simple detection results.

All existing solutions do not allow objective automation.

Disclosure of Invention

In order to solve the technical problems, the invention discloses a pandemic early-stage discovery system and a method based on big data and artificial intelligence, and the technical scheme of the invention is implemented as follows:

an epidemic early discovery system and method based on big data and artificial intelligence, which comprises the following steps: the system comprises an intelligent data acquisition device and a data center; the intelligent data acquisition device acquires and transmits information to the data center; the method comprises the following steps:

s1: collecting survey information, preliminarily analyzing and judging the tendency of the infectious disease, and forming a recommendation conclusion by the collected survey information and a judgment result and transmitting the recommendation conclusion to a data center;

s2: the data center gathers all the obtained suspected syndromes to form a global syndrome list, wherein the corresponding value of the suspected syndromes is a, and the list is formed;

s3: the data center judges whether the signs of the infectious diseases exist or not by using the syndrome data in the steps and constructs an infectious spatio-temporal network by using the spatio-temporal data;

s4: judging the transmission path of the infectious disease through the infection space-time network obtained in the step S3 and carrying out infectious disease inspection and investigation;

wherein, the suspected syndrome discriminating algorithm in S2 includes the following steps:

s2-1: preparing a global syndrome list;

s2-2: taking a symptom data set S _i ＝{s ₁ ,s ₂ ,s ₃ \8230 }, and removing non-infectious disease symptom elements;

s2-3: generation of S _i Power set of { { s { [ ₁ },{s ₂ },{s ₃ },{s ₁ ,s ₂ },{s ₂ ,s ₃ },…}；

S2-4: to S _i All elements in the power set are inquired and compared with a medical knowledge base;

s2-5: all symptom sets meeting epidemic characteristics are reserved to form a suspected syndrome set;

s2-6: checking the syndrome list, checking whether a suspected syndrome exists, if so, adding 1 to the suspected syndrome count a, if not, establishing a new suspected syndrome and adding the new suspected syndrome to the syndrome list;

s2-7: when all data are processed, outputting a syndrome list and a list of corresponding counts a.

The infectivity discrimination algorithm in the S3 comprises the following steps:

S3-1: obtaining a syndrome S _t ；

S3-2: screening everyone for satisfaction of S _t Is S _i When the data is not stored, the data is stored;

s3-3: extracting the screened space-time data of each person, and creating a plurality of data points when the data is not unique;

s3-4: normalizing the time-space data, and mapping all the time-space data to a three-dimensional time-space;

s3-5: selecting a space-time distance d, checking whether each data point has an adjacent point according to the distance radius, and connecting;

s3-6: checking the network diameter D for a network in which a plurality of points are connected or for several independent networks;

s3-7: if D is larger than the preset value, the disease is judged to be infectious, otherwise, the syndrome judgment is ended.

Preferably, the method further comprises the following steps of S2A:

and recording historical infectious disease flow records through a data center, extracting specific symptoms from the data and combining with historical diagnosis results, and training a syndrome recommendation system to assist in constructing a global syndrome list.

Preferably, the method further comprises S2B: sorting the syndrome by adopting an infection factor algorithm;

the infection factor algorithm comprises the following steps:

S2B-1: searching a subset relation in the global syndrome list;

S2B-2: the infectious agent count for one syndrome is b, which is incremented by 1 when one syndrome is a subset of another;

S2B-3: the infectious agent Min (a) was calculated for all syndromes _i ，b _i )；

S2B-4: and outputting a syndrome list according to the sorting order of the infectious agents.

Preferably, the intelligent data acquisition devices comprise medical institutions, community public places, families and personal intelligent data acquisition devices.

Preferably, the data collected by the intelligent data collection device comprises an artificial intelligence inquiry table, a detection result, a recording condition, an analysis result table, time and geographical position information.

Preferably, the method for training the syndrome recommendation system in S2A comprises a Bozmann machine, a deep neural network and a random forest.

Preferably, the steps S3-5 and S3-6 set spatiotemporal distances using a clustering algorithm.

By implementing the technical scheme of the invention, the technical problem that epidemic diseases cannot be objectively and actively investigated and judged in the prior art can be solved; by implementing the technical scheme of the invention, the suspected cases are automatically identified, automatically reported and automatically analyzed by adopting big data and artificial intelligence technology, and the migration of the specific syndrome is taken as a trigger point for judging whether infectious diseases exist, so that the technical effect of providing stable early detection of epidemic diseases can be realized.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only one embodiment of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.

FIG. 1 is a schematic diagram of the system of the present invention;

FIG. 2 is a simplified process flow diagram of the present invention;

FIG. 3 is a flow chart of a suspected syndrome discriminating algorithm;

FIG. 4 is a schematic flow chart of an infectivity discrimination algorithm;

FIG. 5 is a schematic flow chart of an infection factor algorithm;

FIG. 6 is a diagram of normalized spatio-temporal coordinate structure.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

Examples

In a specific embodiment, as shown in fig. 1, fig. 2, fig. 3 and fig. 4, a big data and artificial intelligence based epidemic early discovery system and method is characterized in that: the system comprises an intelligent data acquisition device and a data center; the intelligent data acquisition device acquires and transmits information to the data center; the method comprises the following steps:

s2: the data center collects all the obtained suspected syndromes to form a global syndrome list, and the corresponding value of the suspected syndromes is a to form a list;

the suspected syndrome discrimination algorithm in the step S2 comprises the following steps:

s2-1: preparing a global syndrome list;

s2-3: generating S _i Power set of { { s { [ ₁ },{s ₂ },{s ₃ },{s ₁ ,s ₂ },{s ₂ ,s ₃ },…}；

s2-6: checking the syndrome list, checking whether a suspected syndrome exists, if so, adding 1 to the suspected syndrome count a, and if not, establishing a new suspected syndrome and adding the new suspected syndrome to the syndrome list;

s3-1: obtaining a syndrome S _t ；

S3-2: screening everyone to satisfy S _t If the subset is the subset of Si, the data is reserved, otherwise, the data is ignored;

s3-7: if D is larger than the preset value, judging the disease is infectious, otherwise, ending the syndrome judgment.

In this embodiment, the intelligent data acquisition device can complete artificial intelligence inquiry, detection, recording and analysis, and record time and geographic position, the information can be deleted as required or more information can be selected according to actual conditions, such as the travel condition of personnel, and the like, the result output by the intelligent data acquisition device is data and recommendation conclusion, S _i Is a set of symptoms S _i ＝{s ₁ ,s ₂ ,s ₃ \8230 { { S } whose power set is { { S { (S) } ₁ },{s ₂ },{s ₃ },{s ₁ ,s ₂ },{s ₂ ,s ₃ \8230 } where each element is a set, which may be an empty set, a single-element set, a multiple-element set. Here, S _i Can describe symptoms in a text, or can be a detection result to inquire medical knowledge of the external infectious diseasesThe library, it can be known that element in the power set is a set of symptoms specific to infectious disease. Therefore, S2-2 step can be eliminated _i The steps S2-4 and S2-5 can further eliminate the atypical symptom combinations, such as { silty }, { haemorrhage }, etc., and the remaining symptom sets are possible infectious disease symptoms, such as { high fever, chest distress, severe cough }, which are synthesized into the suspected symptom sets. A global syndrome list is constructed at the entry of the algorithm, and a syndrome is a symptom set but the initial state is an empty list. S2-6, when searching in the global syndrome list, all suspected syndromes output in the S2-5 exist, if yes, the corresponding syndrome count a in the global syndrome list is added with 1, otherwise, the syndrome is added into the global syndrome list. And when all the infectious disease typical symptom sets of all the persons are processed, outputting the obtained global syndrome list and the corresponding numerical value of the counter a. The algorithm can calculate the occurrence frequency of possible infectious disease syndromes in all people, and the greater the value a is, the greater the probability of the occurrence of the infectious disease class corresponding to the syndrome is represented according to the sorting of the value a. The algorithm is suitable for situations where the specific or typical symptoms are indeed clear, i.e. where there is a target infectious disease.

The likelihood that the syndrome listed above is a potential infectious disease is high. But this still does not have a fair confidence to initiate a check-up. The primary characteristic of infectious diseases is infectivity. Therefore, the embodiment also provides an infectivity discrimination algorithm, which judges whether the syndrome has transitivity according to the spatiotemporal data of each person. The algorithm may be repeatedly invoked to process each syndrome in the list of syndromes generated by the suspected syndrome discrimination algorithm. First, step S3-1 obtains a syndrome, i.e., a symptom set is named S _t (ii) a S3-2 step checks Each person' S symptom set data S _i When S is _t Is S _i When the user is in the first sub-set, the data of the current person is kept, otherwise, the data is ignored; s3-3, checking the spatio-temporal data of each person screened in the S3-2, wherein when a plurality of spatio-temporal data pairs exist, namely one person at different timesAnd making multiple data records of different places, and generating multiple time-space data points for the person; s3-4, reasonably normalizing all space-time data, expressing the space data by using a 2-dimensional plane (x, y), using the time data as a z-axis to form a three-dimensional space, and enabling the distance between each unit time and each unit space to be equidistant on a coordinate axis, namely forming a cube by the unit space-time; s3-5, selecting a proper space-time distance d in the three-dimensional space, taking each data point as a center, making a sphere according to the distance as a diameter, and if any two spheres are intersected, connecting the two center points to form a network or a plurality of independent networks; s3-6, checking the maximum network diameter D, and if D is larger than a preset value, judging that the syndrome has the infection characteristic. S3-4, the time and space data need to be normalized, and then the data of the person corresponds to specific time and distance points. The characteristics of the spatio-temporal data acquisition determine that neither is it possible to achieve continuous acquisition or sufficiently dense acquisition. In a specific embodiment, for example, data of 5 persons, positions with a, B, C, D, and E as centers can be determined, time periods with a, B, C, D, and E as centers can be determined, position information does not need to be sorted, and time information needs to be sorted. Because the cost for determining the three-dimensional space position in the actual data acquisition is too high, the space data can be placed on a two-dimensional space-time plane according to the actual numerical value. The projection overlapping of the crowd in the two-dimensional space plane position is easily generated in one building, and in this case, the relative position information such as Bluetooth can judge whether the short-distance contact exists, so that the 3-dimensional space and the 1-dimensional time are simplified into the 3-dimensional space-time space of the 2-dimensional plane and the 1-dimensional time, which is reasonable and simplified approximation. Assuming that the maximum diameter of the positions A to E is m meters, the maximum time span of a to E is n minutes, and the unit time of the time axis is p minutes, the unit distance of the space plane is mn/p. Such normalization may simplify model training within the three-dimensional spatio-temporal space described above.

After the above steps are completed, as shown in fig. 6, a spatial network formed by one or more data clusters is formed in a three-dimensional space-time space, and on the premise of such a data topology, different clusters are connected to form a path basis for epidemiological investigation. It is finally output in text form as a list with information of person ID, location, time, syndrome, etc.

In a preferred embodiment, the method further comprises the following steps of S2A:

and recording historical infectious disease flow records through a data center, extracting specific symptoms from the data, combining the specific symptoms with historical diagnosis results, and training a syndrome recommendation system to assist in constructing a global syndrome list.

The embodiment optimizes the suspected syndrome distinguishing algorithm, and the required operation time may be longer when the data volume is large in practice. At the same time, the assistance of the medical knowledge base of external infectious diseases is required. In case of sufficient historical infectious disease epidemics, such as hundreds of thousands or millions of SARS or new coronary epidemics, dengue epidemic, avian influenza epidemic, etc., specific symptoms can be extracted from these data in combination with historical diagnosis results to train a syndrome recommendation system. The function of finding potential syndrome and sequencing can be achieved. Therefore, the epidemic disease investigation speed can be improved, the calculation amount of the algorithm is saved, the calculation efficiency is improved, and the data center can conveniently judge the data which is in accordance with the traditional infectious diseases in the data information to classify.

In a preferred embodiment, as shown in fig. 5, the method further comprises S2B: sorting the syndrome by adopting an infection factor algorithm;

the infection factor algorithm comprises the following steps:

S2B-1: searching a subset relation in the global syndrome list;

S2B-3: calculating the infectious agent Min (a) for all syndromes _i ，b _i )；

S2B-4: and outputting a syndrome list according to the sequence of the infectious factors.

In this embodiment, the primary purpose of the infectious agent calculation is to be able to lock in one or several potential possibilities when the infectious disease target is unknown. The algorithm is based on the discrimination calculation of suspected syndromeAnd on the basis of the method, obtaining a global syndrome list and an a value. First, a subset relationship is looked up among all syndromes, e.g., { s } ₁ ，s ₂ ，s ₃ Is a syndrome for s ₁ ，s ₂ ，s ₃ ，s ₄ ，s ₅ The former is a subset of the latter. When a subset relation is determined, 1 is added to the former counter b, and so on, a compilation is processed for the whole syndrome. When all treatment is completed, the infectious agent is defined as min { a } _i ，b _i Calculate the infectious agent for each syndrome in the syndrome list. The intuitive meaning of an infectious agent is that the value smaller between the number of people in a symptom group and the number of symptom groups related thereto is the influence of the symptom group as an infectious disease. When all syndromes are ranked by infectious agent, where the larger the infectious agent, the larger the population and disease complex that the set of symptoms covers may be. E.g. { s ₁ ，s ₂ ，s ₃ The symbol is a set of symptoms with a value of 100, b value of 20, the infectious agent of the symptom of 20, and another set of symptoms { s } ₄ ，s ₅ ，s ₆ A is 50, b is 30, and the infectious agent is 30, the probability of the latter should be checked preferentially.

The present embodiment may give priority to screening examinations in the case of unknown target infectious diseases. The most reliable means of infectious disease prevention and judgment is whether a test assay is available, and the present invention can give priority to test assays.

In a preferred embodiment, the intelligent data acquisition devices include medical institutions, community public places, home and personal intelligent data acquisition devices.

The embodiment provides a distribution scheme of intelligent data acquisition devices, which realizes the information acquisition of medical institutions, community public places, families and individuals through the intelligent data acquisition devices of the medical institutions, the community public places, the families and the individuals, and implements the automatic flow regulation of epidemic diseases on the basis of the information acquisition.

In a preferred embodiment, the data collected by the intelligent data collecting device comprises an artificial intelligence inquiry table, a detection result, a recording condition, an analysis result table, time and geographic position information.

This embodiment provides an information acquisition template for an intelligent data acquisition device, which is used to investigate epidemics by acquiring such information.

In a preferred embodiment, the method of S2A training syndrome recommendation system includes a bauzmann machine, a deep neural network, and a random forest.

Some methods for training a syndrome recommendation system are given in this embodiment, but other methods that can achieve the same or similar effects may also be implemented, and may be selected or used together according to the actual situation.

In a preferred embodiment, the steps S3-5 and S3-6 set spatiotemporal distances using a clustering algorithm.

And S3-5 and S3-6, judging the infectivity according to the diameter of the communication network by setting space-time distance parameters. The method can be realized by a clustering algorithm, and particularly, the clustering algorithm based on density has advantages. In the case where the historical data is sufficient and the spatiotemporal data is complete, the historical referenceable parameters may also be determined by statistical methods. The selection is not absolute, and can be selected or used together according to actual conditions.

The invention adopts a medical artificial intelligent inquiry and monitoring system as an intelligent data acquisition device, has different forms for different scenes, and can complete the functions of intelligent symptom collection, life index monitoring and time and position data recording;

the data generated by the intelligent data acquisition device comprises timestamp data, geographical position data, symptom data answered by the user and life index detection data. The data is used for judging whether a certain disease tendency exists or not locally, and meanwhile, the data is sent to a data center, and the data is further processed in the data center to achieve the purpose of judging infectivity.

According to the invention, a series of specific symptoms are determined according to the medical knowledge of infectious diseases; an algorithm specific to the present invention is applied to the population data to analyze whether a specific set of symptoms (hereinafter referred to as syndrome) is transmitted in the population, and if so, to determine that an infectious disease is in transmission. The invention tightly grasps the main characteristic of infectious disease, namely 'infection', to judge whether the disease is transmitted or not, but not directly judge the specific disease, thereby having remarkable effect on infectious diseases, particularly position infectious diseases.

After a specific syndrome is obtained, the individual data of the data center form a large data network according to the association relation of time, space and the syndrome. On the data network, the invention provides a specific algorithm to calculate the most possible propagation path, so as to achieve the effect of intelligent automatic flow regulation.

It should be understood that the above-described embodiments are merely exemplary of the present invention, and are not intended to limit the present invention, and that any modification, equivalent replacement, or improvement made without departing from the spirit and principle of the present invention shall fall within the protection scope of the present invention.

Claims

1. An epidemic early discovery system and method based on big data and artificial intelligence are characterized in that: the system comprises an intelligent data acquisition device and a data center; the intelligent data acquisition device acquires and transmits information to the data center; the method comprises the following steps:

s2-1: preparing a global syndrome list;

s2-3: generating S _i Power set of { { s { (S) } ₁ },{s ₂ },{s ₃ },{s ₁ ,s ₂ },{s ₂ ,s ₃ },…}；

s2-6: checking the syndrome list, checking whether a suspected syndrome exists, if so, adding 1 to the suspected syndrome count a, and if not, adding a current new suspected syndrome and adding the current new suspected syndrome to the syndrome list;

s2-7: judging whether the data is processed or not, if not, returning to S2 for continuous operation, and if so, outputting a syndrome list and a list of corresponding counts a;

s3-1: obtaining a syndrome S _t ；

S3-2: screening everyone to satisfy S _t Is S _i When the data is to be collected, the data is stored, otherwise, the data is ignored;

s3-3: extracting the screened spatiotemporal data of each person, and creating a plurality of data points when the data is not unique;

s3-4: normalizing the space-time data, and mapping all the space-time data to a three-dimensional space-time space;

2. The system and method for epidemic early detection based on big data and artificial intelligence of claim 1, wherein: also includes S2A: and recording historical infectious disease flow records through a data center, extracting specific symptoms from the data and combining with historical diagnosis results, and training a syndrome recommendation system to assist in constructing a global syndrome list.

3. The system and method for epidemic early detection based on big data and artificial intelligence as claimed in claim 2, wherein: further comprising S2B: sorting the syndrome by adopting an infection factor algorithm;

the infection factor algorithm comprises the following steps:

S2B-1: searching a subset relation in the global syndrome list;

4. The system and method for epidemic early detection based on big data and artificial intelligence of claim 3, wherein: the intelligent data acquisition device comprises medical institutions, community public places, families and personal intelligent data acquisition devices.

5. The epidemic early discovery system and method based on big data and artificial intelligence of claim 4, wherein: the data collected by the intelligent data collecting device comprises an artificial intelligent inquiry table, a detection result, a recording condition, an analysis result table, time and geographical position information.

6. The system and method for early epidemic detection based on big data and artificial intelligence of claim 5, wherein: the S2A training syndrome recommendation system method comprises a Botzmann machine, a deep neural network and a random forest.

7. The system and method for epidemic early detection based on big data and artificial intelligence as claimed in claim 6, wherein: and the step S3-5 and the step S3-6 use a clustering algorithm to set the space-time distance.